2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
44 #include "../nb_kernel.h"
45 #include "gromacs/legacyheaders/types/simple.h"
46 #include "gromacs/math/vec.h"
47 #include "gromacs/legacyheaders/nrnb.h"
49 #include "gromacs/simd/math_x86_sse2_double.h"
50 #include "kernelutil_x86_sse2_double.h"
53 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW4W4_VF_sse2_double
54 * Electrostatics interaction: Ewald
55 * VdW interaction: LJEwald
56 * Geometry: Water4-Water4
57 * Calculate force/pot: PotentialAndForce
60 nb_kernel_ElecEw_VdwLJEw_GeomW4W4_VF_sse2_double
61 (t_nblist * gmx_restrict nlist,
62 rvec * gmx_restrict xx,
63 rvec * gmx_restrict ff,
64 t_forcerec * gmx_restrict fr,
65 t_mdatoms * gmx_restrict mdatoms,
66 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
67 t_nrnb * gmx_restrict nrnb)
69 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
70 * just 0 for non-waters.
71 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
72 * jnr indices corresponding to data put in the four positions in the SIMD register.
74 int i_shift_offset,i_coord_offset,outeriter,inneriter;
75 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
77 int j_coord_offsetA,j_coord_offsetB;
78 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
80 real *shiftvec,*fshift,*x,*f;
81 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
83 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
85 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
87 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
89 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
90 int vdwjidx0A,vdwjidx0B;
91 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
92 int vdwjidx1A,vdwjidx1B;
93 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
94 int vdwjidx2A,vdwjidx2B;
95 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
96 int vdwjidx3A,vdwjidx3B;
97 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
98 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
99 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
100 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
101 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
102 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
103 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
104 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
105 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
106 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
107 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
108 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
111 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
114 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
115 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
126 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
128 __m128d one_half = _mm_set1_pd(0.5);
129 __m128d minus_one = _mm_set1_pd(-1.0);
131 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
133 __m128d dummy_mask,cutoff_mask;
134 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
135 __m128d one = _mm_set1_pd(1.0);
136 __m128d two = _mm_set1_pd(2.0);
142 jindex = nlist->jindex;
144 shiftidx = nlist->shift;
146 shiftvec = fr->shift_vec[0];
147 fshift = fr->fshift[0];
148 facel = _mm_set1_pd(fr->epsfac);
149 charge = mdatoms->chargeA;
150 nvdwtype = fr->ntype;
152 vdwtype = mdatoms->typeA;
153 vdwgridparam = fr->ljpme_c6grid;
154 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
155 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
156 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
158 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
159 ewtab = fr->ic->tabq_coul_FDV0;
160 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
161 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
163 /* Setup water-specific parameters */
164 inr = nlist->iinr[0];
165 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
166 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
167 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
168 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
170 jq1 = _mm_set1_pd(charge[inr+1]);
171 jq2 = _mm_set1_pd(charge[inr+2]);
172 jq3 = _mm_set1_pd(charge[inr+3]);
173 vdwjidx0A = 2*vdwtype[inr+0];
174 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
175 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
176 c6grid_00 = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
177 qq11 = _mm_mul_pd(iq1,jq1);
178 qq12 = _mm_mul_pd(iq1,jq2);
179 qq13 = _mm_mul_pd(iq1,jq3);
180 qq21 = _mm_mul_pd(iq2,jq1);
181 qq22 = _mm_mul_pd(iq2,jq2);
182 qq23 = _mm_mul_pd(iq2,jq3);
183 qq31 = _mm_mul_pd(iq3,jq1);
184 qq32 = _mm_mul_pd(iq3,jq2);
185 qq33 = _mm_mul_pd(iq3,jq3);
187 /* Avoid stupid compiler warnings */
195 /* Start outer loop over neighborlists */
196 for(iidx=0; iidx<nri; iidx++)
198 /* Load shift vector for this list */
199 i_shift_offset = DIM*shiftidx[iidx];
201 /* Load limits for loop over neighbors */
202 j_index_start = jindex[iidx];
203 j_index_end = jindex[iidx+1];
205 /* Get outer coordinate index */
207 i_coord_offset = DIM*inr;
209 /* Load i particle coords and add shift vector */
210 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
211 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
213 fix0 = _mm_setzero_pd();
214 fiy0 = _mm_setzero_pd();
215 fiz0 = _mm_setzero_pd();
216 fix1 = _mm_setzero_pd();
217 fiy1 = _mm_setzero_pd();
218 fiz1 = _mm_setzero_pd();
219 fix2 = _mm_setzero_pd();
220 fiy2 = _mm_setzero_pd();
221 fiz2 = _mm_setzero_pd();
222 fix3 = _mm_setzero_pd();
223 fiy3 = _mm_setzero_pd();
224 fiz3 = _mm_setzero_pd();
226 /* Reset potential sums */
227 velecsum = _mm_setzero_pd();
228 vvdwsum = _mm_setzero_pd();
230 /* Start inner kernel loop */
231 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
234 /* Get j neighbor index, and coordinate index */
237 j_coord_offsetA = DIM*jnrA;
238 j_coord_offsetB = DIM*jnrB;
240 /* load j atom coordinates */
241 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
242 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
243 &jy2,&jz2,&jx3,&jy3,&jz3);
245 /* Calculate displacement vector */
246 dx00 = _mm_sub_pd(ix0,jx0);
247 dy00 = _mm_sub_pd(iy0,jy0);
248 dz00 = _mm_sub_pd(iz0,jz0);
249 dx11 = _mm_sub_pd(ix1,jx1);
250 dy11 = _mm_sub_pd(iy1,jy1);
251 dz11 = _mm_sub_pd(iz1,jz1);
252 dx12 = _mm_sub_pd(ix1,jx2);
253 dy12 = _mm_sub_pd(iy1,jy2);
254 dz12 = _mm_sub_pd(iz1,jz2);
255 dx13 = _mm_sub_pd(ix1,jx3);
256 dy13 = _mm_sub_pd(iy1,jy3);
257 dz13 = _mm_sub_pd(iz1,jz3);
258 dx21 = _mm_sub_pd(ix2,jx1);
259 dy21 = _mm_sub_pd(iy2,jy1);
260 dz21 = _mm_sub_pd(iz2,jz1);
261 dx22 = _mm_sub_pd(ix2,jx2);
262 dy22 = _mm_sub_pd(iy2,jy2);
263 dz22 = _mm_sub_pd(iz2,jz2);
264 dx23 = _mm_sub_pd(ix2,jx3);
265 dy23 = _mm_sub_pd(iy2,jy3);
266 dz23 = _mm_sub_pd(iz2,jz3);
267 dx31 = _mm_sub_pd(ix3,jx1);
268 dy31 = _mm_sub_pd(iy3,jy1);
269 dz31 = _mm_sub_pd(iz3,jz1);
270 dx32 = _mm_sub_pd(ix3,jx2);
271 dy32 = _mm_sub_pd(iy3,jy2);
272 dz32 = _mm_sub_pd(iz3,jz2);
273 dx33 = _mm_sub_pd(ix3,jx3);
274 dy33 = _mm_sub_pd(iy3,jy3);
275 dz33 = _mm_sub_pd(iz3,jz3);
277 /* Calculate squared distance and things based on it */
278 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
279 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
280 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
281 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
282 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
283 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
284 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
285 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
286 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
287 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
289 rinv00 = gmx_mm_invsqrt_pd(rsq00);
290 rinv11 = gmx_mm_invsqrt_pd(rsq11);
291 rinv12 = gmx_mm_invsqrt_pd(rsq12);
292 rinv13 = gmx_mm_invsqrt_pd(rsq13);
293 rinv21 = gmx_mm_invsqrt_pd(rsq21);
294 rinv22 = gmx_mm_invsqrt_pd(rsq22);
295 rinv23 = gmx_mm_invsqrt_pd(rsq23);
296 rinv31 = gmx_mm_invsqrt_pd(rsq31);
297 rinv32 = gmx_mm_invsqrt_pd(rsq32);
298 rinv33 = gmx_mm_invsqrt_pd(rsq33);
300 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
301 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
302 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
303 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
304 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
305 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
306 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
307 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
308 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
309 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
311 fjx0 = _mm_setzero_pd();
312 fjy0 = _mm_setzero_pd();
313 fjz0 = _mm_setzero_pd();
314 fjx1 = _mm_setzero_pd();
315 fjy1 = _mm_setzero_pd();
316 fjz1 = _mm_setzero_pd();
317 fjx2 = _mm_setzero_pd();
318 fjy2 = _mm_setzero_pd();
319 fjz2 = _mm_setzero_pd();
320 fjx3 = _mm_setzero_pd();
321 fjy3 = _mm_setzero_pd();
322 fjz3 = _mm_setzero_pd();
324 /**************************
325 * CALCULATE INTERACTIONS *
326 **************************/
328 r00 = _mm_mul_pd(rsq00,rinv00);
330 /* Analytical LJ-PME */
331 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
332 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
333 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
334 exponent = gmx_simd_exp_d(ewcljrsq);
335 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
336 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
337 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
338 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
339 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
340 vvdw = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
341 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
342 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
344 /* Update potential sum for this i atom from the interaction with this j atom. */
345 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
349 /* Calculate temporary vectorial force */
350 tx = _mm_mul_pd(fscal,dx00);
351 ty = _mm_mul_pd(fscal,dy00);
352 tz = _mm_mul_pd(fscal,dz00);
354 /* Update vectorial force */
355 fix0 = _mm_add_pd(fix0,tx);
356 fiy0 = _mm_add_pd(fiy0,ty);
357 fiz0 = _mm_add_pd(fiz0,tz);
359 fjx0 = _mm_add_pd(fjx0,tx);
360 fjy0 = _mm_add_pd(fjy0,ty);
361 fjz0 = _mm_add_pd(fjz0,tz);
363 /**************************
364 * CALCULATE INTERACTIONS *
365 **************************/
367 r11 = _mm_mul_pd(rsq11,rinv11);
369 /* EWALD ELECTROSTATICS */
371 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
372 ewrt = _mm_mul_pd(r11,ewtabscale);
373 ewitab = _mm_cvttpd_epi32(ewrt);
374 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
375 ewitab = _mm_slli_epi32(ewitab,2);
376 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
377 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
378 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
379 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
380 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
381 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
382 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
383 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
384 velec = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
385 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
387 /* Update potential sum for this i atom from the interaction with this j atom. */
388 velecsum = _mm_add_pd(velecsum,velec);
392 /* Calculate temporary vectorial force */
393 tx = _mm_mul_pd(fscal,dx11);
394 ty = _mm_mul_pd(fscal,dy11);
395 tz = _mm_mul_pd(fscal,dz11);
397 /* Update vectorial force */
398 fix1 = _mm_add_pd(fix1,tx);
399 fiy1 = _mm_add_pd(fiy1,ty);
400 fiz1 = _mm_add_pd(fiz1,tz);
402 fjx1 = _mm_add_pd(fjx1,tx);
403 fjy1 = _mm_add_pd(fjy1,ty);
404 fjz1 = _mm_add_pd(fjz1,tz);
406 /**************************
407 * CALCULATE INTERACTIONS *
408 **************************/
410 r12 = _mm_mul_pd(rsq12,rinv12);
412 /* EWALD ELECTROSTATICS */
414 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
415 ewrt = _mm_mul_pd(r12,ewtabscale);
416 ewitab = _mm_cvttpd_epi32(ewrt);
417 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
418 ewitab = _mm_slli_epi32(ewitab,2);
419 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
420 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
421 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
422 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
423 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
424 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
425 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
426 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
427 velec = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
428 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
430 /* Update potential sum for this i atom from the interaction with this j atom. */
431 velecsum = _mm_add_pd(velecsum,velec);
435 /* Calculate temporary vectorial force */
436 tx = _mm_mul_pd(fscal,dx12);
437 ty = _mm_mul_pd(fscal,dy12);
438 tz = _mm_mul_pd(fscal,dz12);
440 /* Update vectorial force */
441 fix1 = _mm_add_pd(fix1,tx);
442 fiy1 = _mm_add_pd(fiy1,ty);
443 fiz1 = _mm_add_pd(fiz1,tz);
445 fjx2 = _mm_add_pd(fjx2,tx);
446 fjy2 = _mm_add_pd(fjy2,ty);
447 fjz2 = _mm_add_pd(fjz2,tz);
449 /**************************
450 * CALCULATE INTERACTIONS *
451 **************************/
453 r13 = _mm_mul_pd(rsq13,rinv13);
455 /* EWALD ELECTROSTATICS */
457 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
458 ewrt = _mm_mul_pd(r13,ewtabscale);
459 ewitab = _mm_cvttpd_epi32(ewrt);
460 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
461 ewitab = _mm_slli_epi32(ewitab,2);
462 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
463 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
464 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
465 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
466 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
467 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
468 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
469 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
470 velec = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
471 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
473 /* Update potential sum for this i atom from the interaction with this j atom. */
474 velecsum = _mm_add_pd(velecsum,velec);
478 /* Calculate temporary vectorial force */
479 tx = _mm_mul_pd(fscal,dx13);
480 ty = _mm_mul_pd(fscal,dy13);
481 tz = _mm_mul_pd(fscal,dz13);
483 /* Update vectorial force */
484 fix1 = _mm_add_pd(fix1,tx);
485 fiy1 = _mm_add_pd(fiy1,ty);
486 fiz1 = _mm_add_pd(fiz1,tz);
488 fjx3 = _mm_add_pd(fjx3,tx);
489 fjy3 = _mm_add_pd(fjy3,ty);
490 fjz3 = _mm_add_pd(fjz3,tz);
492 /**************************
493 * CALCULATE INTERACTIONS *
494 **************************/
496 r21 = _mm_mul_pd(rsq21,rinv21);
498 /* EWALD ELECTROSTATICS */
500 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
501 ewrt = _mm_mul_pd(r21,ewtabscale);
502 ewitab = _mm_cvttpd_epi32(ewrt);
503 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
504 ewitab = _mm_slli_epi32(ewitab,2);
505 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
506 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
507 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
508 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
509 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
510 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
511 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
512 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
513 velec = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
514 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
516 /* Update potential sum for this i atom from the interaction with this j atom. */
517 velecsum = _mm_add_pd(velecsum,velec);
521 /* Calculate temporary vectorial force */
522 tx = _mm_mul_pd(fscal,dx21);
523 ty = _mm_mul_pd(fscal,dy21);
524 tz = _mm_mul_pd(fscal,dz21);
526 /* Update vectorial force */
527 fix2 = _mm_add_pd(fix2,tx);
528 fiy2 = _mm_add_pd(fiy2,ty);
529 fiz2 = _mm_add_pd(fiz2,tz);
531 fjx1 = _mm_add_pd(fjx1,tx);
532 fjy1 = _mm_add_pd(fjy1,ty);
533 fjz1 = _mm_add_pd(fjz1,tz);
535 /**************************
536 * CALCULATE INTERACTIONS *
537 **************************/
539 r22 = _mm_mul_pd(rsq22,rinv22);
541 /* EWALD ELECTROSTATICS */
543 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
544 ewrt = _mm_mul_pd(r22,ewtabscale);
545 ewitab = _mm_cvttpd_epi32(ewrt);
546 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
547 ewitab = _mm_slli_epi32(ewitab,2);
548 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
549 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
550 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
551 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
552 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
553 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
554 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
555 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
556 velec = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
557 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
559 /* Update potential sum for this i atom from the interaction with this j atom. */
560 velecsum = _mm_add_pd(velecsum,velec);
564 /* Calculate temporary vectorial force */
565 tx = _mm_mul_pd(fscal,dx22);
566 ty = _mm_mul_pd(fscal,dy22);
567 tz = _mm_mul_pd(fscal,dz22);
569 /* Update vectorial force */
570 fix2 = _mm_add_pd(fix2,tx);
571 fiy2 = _mm_add_pd(fiy2,ty);
572 fiz2 = _mm_add_pd(fiz2,tz);
574 fjx2 = _mm_add_pd(fjx2,tx);
575 fjy2 = _mm_add_pd(fjy2,ty);
576 fjz2 = _mm_add_pd(fjz2,tz);
578 /**************************
579 * CALCULATE INTERACTIONS *
580 **************************/
582 r23 = _mm_mul_pd(rsq23,rinv23);
584 /* EWALD ELECTROSTATICS */
586 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
587 ewrt = _mm_mul_pd(r23,ewtabscale);
588 ewitab = _mm_cvttpd_epi32(ewrt);
589 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
590 ewitab = _mm_slli_epi32(ewitab,2);
591 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
592 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
593 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
594 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
595 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
596 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
597 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
598 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
599 velec = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
600 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
602 /* Update potential sum for this i atom from the interaction with this j atom. */
603 velecsum = _mm_add_pd(velecsum,velec);
607 /* Calculate temporary vectorial force */
608 tx = _mm_mul_pd(fscal,dx23);
609 ty = _mm_mul_pd(fscal,dy23);
610 tz = _mm_mul_pd(fscal,dz23);
612 /* Update vectorial force */
613 fix2 = _mm_add_pd(fix2,tx);
614 fiy2 = _mm_add_pd(fiy2,ty);
615 fiz2 = _mm_add_pd(fiz2,tz);
617 fjx3 = _mm_add_pd(fjx3,tx);
618 fjy3 = _mm_add_pd(fjy3,ty);
619 fjz3 = _mm_add_pd(fjz3,tz);
621 /**************************
622 * CALCULATE INTERACTIONS *
623 **************************/
625 r31 = _mm_mul_pd(rsq31,rinv31);
627 /* EWALD ELECTROSTATICS */
629 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
630 ewrt = _mm_mul_pd(r31,ewtabscale);
631 ewitab = _mm_cvttpd_epi32(ewrt);
632 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
633 ewitab = _mm_slli_epi32(ewitab,2);
634 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
635 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
636 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
637 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
638 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
639 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
640 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
641 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
642 velec = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
643 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
645 /* Update potential sum for this i atom from the interaction with this j atom. */
646 velecsum = _mm_add_pd(velecsum,velec);
650 /* Calculate temporary vectorial force */
651 tx = _mm_mul_pd(fscal,dx31);
652 ty = _mm_mul_pd(fscal,dy31);
653 tz = _mm_mul_pd(fscal,dz31);
655 /* Update vectorial force */
656 fix3 = _mm_add_pd(fix3,tx);
657 fiy3 = _mm_add_pd(fiy3,ty);
658 fiz3 = _mm_add_pd(fiz3,tz);
660 fjx1 = _mm_add_pd(fjx1,tx);
661 fjy1 = _mm_add_pd(fjy1,ty);
662 fjz1 = _mm_add_pd(fjz1,tz);
664 /**************************
665 * CALCULATE INTERACTIONS *
666 **************************/
668 r32 = _mm_mul_pd(rsq32,rinv32);
670 /* EWALD ELECTROSTATICS */
672 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
673 ewrt = _mm_mul_pd(r32,ewtabscale);
674 ewitab = _mm_cvttpd_epi32(ewrt);
675 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
676 ewitab = _mm_slli_epi32(ewitab,2);
677 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
678 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
679 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
680 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
681 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
682 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
683 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
684 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
685 velec = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
686 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
688 /* Update potential sum for this i atom from the interaction with this j atom. */
689 velecsum = _mm_add_pd(velecsum,velec);
693 /* Calculate temporary vectorial force */
694 tx = _mm_mul_pd(fscal,dx32);
695 ty = _mm_mul_pd(fscal,dy32);
696 tz = _mm_mul_pd(fscal,dz32);
698 /* Update vectorial force */
699 fix3 = _mm_add_pd(fix3,tx);
700 fiy3 = _mm_add_pd(fiy3,ty);
701 fiz3 = _mm_add_pd(fiz3,tz);
703 fjx2 = _mm_add_pd(fjx2,tx);
704 fjy2 = _mm_add_pd(fjy2,ty);
705 fjz2 = _mm_add_pd(fjz2,tz);
707 /**************************
708 * CALCULATE INTERACTIONS *
709 **************************/
711 r33 = _mm_mul_pd(rsq33,rinv33);
713 /* EWALD ELECTROSTATICS */
715 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
716 ewrt = _mm_mul_pd(r33,ewtabscale);
717 ewitab = _mm_cvttpd_epi32(ewrt);
718 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
719 ewitab = _mm_slli_epi32(ewitab,2);
720 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
721 ewtabD = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,1) );
722 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
723 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
724 ewtabFn = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,1) +2);
725 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
726 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
727 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
728 velec = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
729 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
731 /* Update potential sum for this i atom from the interaction with this j atom. */
732 velecsum = _mm_add_pd(velecsum,velec);
736 /* Calculate temporary vectorial force */
737 tx = _mm_mul_pd(fscal,dx33);
738 ty = _mm_mul_pd(fscal,dy33);
739 tz = _mm_mul_pd(fscal,dz33);
741 /* Update vectorial force */
742 fix3 = _mm_add_pd(fix3,tx);
743 fiy3 = _mm_add_pd(fiy3,ty);
744 fiz3 = _mm_add_pd(fiz3,tz);
746 fjx3 = _mm_add_pd(fjx3,tx);
747 fjy3 = _mm_add_pd(fjy3,ty);
748 fjz3 = _mm_add_pd(fjz3,tz);
750 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
752 /* Inner loop uses 423 flops */
759 j_coord_offsetA = DIM*jnrA;
761 /* load j atom coordinates */
762 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
763 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
764 &jy2,&jz2,&jx3,&jy3,&jz3);
766 /* Calculate displacement vector */
767 dx00 = _mm_sub_pd(ix0,jx0);
768 dy00 = _mm_sub_pd(iy0,jy0);
769 dz00 = _mm_sub_pd(iz0,jz0);
770 dx11 = _mm_sub_pd(ix1,jx1);
771 dy11 = _mm_sub_pd(iy1,jy1);
772 dz11 = _mm_sub_pd(iz1,jz1);
773 dx12 = _mm_sub_pd(ix1,jx2);
774 dy12 = _mm_sub_pd(iy1,jy2);
775 dz12 = _mm_sub_pd(iz1,jz2);
776 dx13 = _mm_sub_pd(ix1,jx3);
777 dy13 = _mm_sub_pd(iy1,jy3);
778 dz13 = _mm_sub_pd(iz1,jz3);
779 dx21 = _mm_sub_pd(ix2,jx1);
780 dy21 = _mm_sub_pd(iy2,jy1);
781 dz21 = _mm_sub_pd(iz2,jz1);
782 dx22 = _mm_sub_pd(ix2,jx2);
783 dy22 = _mm_sub_pd(iy2,jy2);
784 dz22 = _mm_sub_pd(iz2,jz2);
785 dx23 = _mm_sub_pd(ix2,jx3);
786 dy23 = _mm_sub_pd(iy2,jy3);
787 dz23 = _mm_sub_pd(iz2,jz3);
788 dx31 = _mm_sub_pd(ix3,jx1);
789 dy31 = _mm_sub_pd(iy3,jy1);
790 dz31 = _mm_sub_pd(iz3,jz1);
791 dx32 = _mm_sub_pd(ix3,jx2);
792 dy32 = _mm_sub_pd(iy3,jy2);
793 dz32 = _mm_sub_pd(iz3,jz2);
794 dx33 = _mm_sub_pd(ix3,jx3);
795 dy33 = _mm_sub_pd(iy3,jy3);
796 dz33 = _mm_sub_pd(iz3,jz3);
798 /* Calculate squared distance and things based on it */
799 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
800 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
801 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
802 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
803 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
804 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
805 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
806 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
807 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
808 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
810 rinv00 = gmx_mm_invsqrt_pd(rsq00);
811 rinv11 = gmx_mm_invsqrt_pd(rsq11);
812 rinv12 = gmx_mm_invsqrt_pd(rsq12);
813 rinv13 = gmx_mm_invsqrt_pd(rsq13);
814 rinv21 = gmx_mm_invsqrt_pd(rsq21);
815 rinv22 = gmx_mm_invsqrt_pd(rsq22);
816 rinv23 = gmx_mm_invsqrt_pd(rsq23);
817 rinv31 = gmx_mm_invsqrt_pd(rsq31);
818 rinv32 = gmx_mm_invsqrt_pd(rsq32);
819 rinv33 = gmx_mm_invsqrt_pd(rsq33);
821 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
822 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
823 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
824 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
825 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
826 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
827 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
828 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
829 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
830 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
832 fjx0 = _mm_setzero_pd();
833 fjy0 = _mm_setzero_pd();
834 fjz0 = _mm_setzero_pd();
835 fjx1 = _mm_setzero_pd();
836 fjy1 = _mm_setzero_pd();
837 fjz1 = _mm_setzero_pd();
838 fjx2 = _mm_setzero_pd();
839 fjy2 = _mm_setzero_pd();
840 fjz2 = _mm_setzero_pd();
841 fjx3 = _mm_setzero_pd();
842 fjy3 = _mm_setzero_pd();
843 fjz3 = _mm_setzero_pd();
845 /**************************
846 * CALCULATE INTERACTIONS *
847 **************************/
849 r00 = _mm_mul_pd(rsq00,rinv00);
851 /* Analytical LJ-PME */
852 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
853 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
854 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
855 exponent = gmx_simd_exp_d(ewcljrsq);
856 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
857 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
858 /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
859 vvdw6 = _mm_mul_pd(_mm_sub_pd(c6_00,_mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly))),rinvsix);
860 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
861 vvdw = _mm_sub_pd(_mm_mul_pd(vvdw12,one_twelfth),_mm_mul_pd(vvdw6,one_sixth));
862 /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
863 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,_mm_sub_pd(vvdw6,_mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6)))),rinvsq00);
865 /* Update potential sum for this i atom from the interaction with this j atom. */
866 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
867 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
871 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
873 /* Calculate temporary vectorial force */
874 tx = _mm_mul_pd(fscal,dx00);
875 ty = _mm_mul_pd(fscal,dy00);
876 tz = _mm_mul_pd(fscal,dz00);
878 /* Update vectorial force */
879 fix0 = _mm_add_pd(fix0,tx);
880 fiy0 = _mm_add_pd(fiy0,ty);
881 fiz0 = _mm_add_pd(fiz0,tz);
883 fjx0 = _mm_add_pd(fjx0,tx);
884 fjy0 = _mm_add_pd(fjy0,ty);
885 fjz0 = _mm_add_pd(fjz0,tz);
887 /**************************
888 * CALCULATE INTERACTIONS *
889 **************************/
891 r11 = _mm_mul_pd(rsq11,rinv11);
893 /* EWALD ELECTROSTATICS */
895 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
896 ewrt = _mm_mul_pd(r11,ewtabscale);
897 ewitab = _mm_cvttpd_epi32(ewrt);
898 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
899 ewitab = _mm_slli_epi32(ewitab,2);
900 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
901 ewtabD = _mm_setzero_pd();
902 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
903 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
904 ewtabFn = _mm_setzero_pd();
905 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
906 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
907 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
908 velec = _mm_mul_pd(qq11,_mm_sub_pd(rinv11,velec));
909 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
911 /* Update potential sum for this i atom from the interaction with this j atom. */
912 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
913 velecsum = _mm_add_pd(velecsum,velec);
917 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
919 /* Calculate temporary vectorial force */
920 tx = _mm_mul_pd(fscal,dx11);
921 ty = _mm_mul_pd(fscal,dy11);
922 tz = _mm_mul_pd(fscal,dz11);
924 /* Update vectorial force */
925 fix1 = _mm_add_pd(fix1,tx);
926 fiy1 = _mm_add_pd(fiy1,ty);
927 fiz1 = _mm_add_pd(fiz1,tz);
929 fjx1 = _mm_add_pd(fjx1,tx);
930 fjy1 = _mm_add_pd(fjy1,ty);
931 fjz1 = _mm_add_pd(fjz1,tz);
933 /**************************
934 * CALCULATE INTERACTIONS *
935 **************************/
937 r12 = _mm_mul_pd(rsq12,rinv12);
939 /* EWALD ELECTROSTATICS */
941 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
942 ewrt = _mm_mul_pd(r12,ewtabscale);
943 ewitab = _mm_cvttpd_epi32(ewrt);
944 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
945 ewitab = _mm_slli_epi32(ewitab,2);
946 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
947 ewtabD = _mm_setzero_pd();
948 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
949 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
950 ewtabFn = _mm_setzero_pd();
951 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
952 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
953 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
954 velec = _mm_mul_pd(qq12,_mm_sub_pd(rinv12,velec));
955 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
957 /* Update potential sum for this i atom from the interaction with this j atom. */
958 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
959 velecsum = _mm_add_pd(velecsum,velec);
963 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
965 /* Calculate temporary vectorial force */
966 tx = _mm_mul_pd(fscal,dx12);
967 ty = _mm_mul_pd(fscal,dy12);
968 tz = _mm_mul_pd(fscal,dz12);
970 /* Update vectorial force */
971 fix1 = _mm_add_pd(fix1,tx);
972 fiy1 = _mm_add_pd(fiy1,ty);
973 fiz1 = _mm_add_pd(fiz1,tz);
975 fjx2 = _mm_add_pd(fjx2,tx);
976 fjy2 = _mm_add_pd(fjy2,ty);
977 fjz2 = _mm_add_pd(fjz2,tz);
979 /**************************
980 * CALCULATE INTERACTIONS *
981 **************************/
983 r13 = _mm_mul_pd(rsq13,rinv13);
985 /* EWALD ELECTROSTATICS */
987 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
988 ewrt = _mm_mul_pd(r13,ewtabscale);
989 ewitab = _mm_cvttpd_epi32(ewrt);
990 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
991 ewitab = _mm_slli_epi32(ewitab,2);
992 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
993 ewtabD = _mm_setzero_pd();
994 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
995 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
996 ewtabFn = _mm_setzero_pd();
997 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
998 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
999 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1000 velec = _mm_mul_pd(qq13,_mm_sub_pd(rinv13,velec));
1001 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1003 /* Update potential sum for this i atom from the interaction with this j atom. */
1004 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1005 velecsum = _mm_add_pd(velecsum,velec);
1009 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1011 /* Calculate temporary vectorial force */
1012 tx = _mm_mul_pd(fscal,dx13);
1013 ty = _mm_mul_pd(fscal,dy13);
1014 tz = _mm_mul_pd(fscal,dz13);
1016 /* Update vectorial force */
1017 fix1 = _mm_add_pd(fix1,tx);
1018 fiy1 = _mm_add_pd(fiy1,ty);
1019 fiz1 = _mm_add_pd(fiz1,tz);
1021 fjx3 = _mm_add_pd(fjx3,tx);
1022 fjy3 = _mm_add_pd(fjy3,ty);
1023 fjz3 = _mm_add_pd(fjz3,tz);
1025 /**************************
1026 * CALCULATE INTERACTIONS *
1027 **************************/
1029 r21 = _mm_mul_pd(rsq21,rinv21);
1031 /* EWALD ELECTROSTATICS */
1033 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1034 ewrt = _mm_mul_pd(r21,ewtabscale);
1035 ewitab = _mm_cvttpd_epi32(ewrt);
1036 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1037 ewitab = _mm_slli_epi32(ewitab,2);
1038 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1039 ewtabD = _mm_setzero_pd();
1040 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1041 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1042 ewtabFn = _mm_setzero_pd();
1043 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1044 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1045 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1046 velec = _mm_mul_pd(qq21,_mm_sub_pd(rinv21,velec));
1047 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1049 /* Update potential sum for this i atom from the interaction with this j atom. */
1050 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1051 velecsum = _mm_add_pd(velecsum,velec);
1055 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1057 /* Calculate temporary vectorial force */
1058 tx = _mm_mul_pd(fscal,dx21);
1059 ty = _mm_mul_pd(fscal,dy21);
1060 tz = _mm_mul_pd(fscal,dz21);
1062 /* Update vectorial force */
1063 fix2 = _mm_add_pd(fix2,tx);
1064 fiy2 = _mm_add_pd(fiy2,ty);
1065 fiz2 = _mm_add_pd(fiz2,tz);
1067 fjx1 = _mm_add_pd(fjx1,tx);
1068 fjy1 = _mm_add_pd(fjy1,ty);
1069 fjz1 = _mm_add_pd(fjz1,tz);
1071 /**************************
1072 * CALCULATE INTERACTIONS *
1073 **************************/
1075 r22 = _mm_mul_pd(rsq22,rinv22);
1077 /* EWALD ELECTROSTATICS */
1079 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1080 ewrt = _mm_mul_pd(r22,ewtabscale);
1081 ewitab = _mm_cvttpd_epi32(ewrt);
1082 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1083 ewitab = _mm_slli_epi32(ewitab,2);
1084 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1085 ewtabD = _mm_setzero_pd();
1086 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1087 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1088 ewtabFn = _mm_setzero_pd();
1089 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1090 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1091 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1092 velec = _mm_mul_pd(qq22,_mm_sub_pd(rinv22,velec));
1093 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1095 /* Update potential sum for this i atom from the interaction with this j atom. */
1096 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1097 velecsum = _mm_add_pd(velecsum,velec);
1101 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1103 /* Calculate temporary vectorial force */
1104 tx = _mm_mul_pd(fscal,dx22);
1105 ty = _mm_mul_pd(fscal,dy22);
1106 tz = _mm_mul_pd(fscal,dz22);
1108 /* Update vectorial force */
1109 fix2 = _mm_add_pd(fix2,tx);
1110 fiy2 = _mm_add_pd(fiy2,ty);
1111 fiz2 = _mm_add_pd(fiz2,tz);
1113 fjx2 = _mm_add_pd(fjx2,tx);
1114 fjy2 = _mm_add_pd(fjy2,ty);
1115 fjz2 = _mm_add_pd(fjz2,tz);
1117 /**************************
1118 * CALCULATE INTERACTIONS *
1119 **************************/
1121 r23 = _mm_mul_pd(rsq23,rinv23);
1123 /* EWALD ELECTROSTATICS */
1125 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1126 ewrt = _mm_mul_pd(r23,ewtabscale);
1127 ewitab = _mm_cvttpd_epi32(ewrt);
1128 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1129 ewitab = _mm_slli_epi32(ewitab,2);
1130 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1131 ewtabD = _mm_setzero_pd();
1132 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1133 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1134 ewtabFn = _mm_setzero_pd();
1135 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1136 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1137 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1138 velec = _mm_mul_pd(qq23,_mm_sub_pd(rinv23,velec));
1139 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1141 /* Update potential sum for this i atom from the interaction with this j atom. */
1142 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1143 velecsum = _mm_add_pd(velecsum,velec);
1147 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1149 /* Calculate temporary vectorial force */
1150 tx = _mm_mul_pd(fscal,dx23);
1151 ty = _mm_mul_pd(fscal,dy23);
1152 tz = _mm_mul_pd(fscal,dz23);
1154 /* Update vectorial force */
1155 fix2 = _mm_add_pd(fix2,tx);
1156 fiy2 = _mm_add_pd(fiy2,ty);
1157 fiz2 = _mm_add_pd(fiz2,tz);
1159 fjx3 = _mm_add_pd(fjx3,tx);
1160 fjy3 = _mm_add_pd(fjy3,ty);
1161 fjz3 = _mm_add_pd(fjz3,tz);
1163 /**************************
1164 * CALCULATE INTERACTIONS *
1165 **************************/
1167 r31 = _mm_mul_pd(rsq31,rinv31);
1169 /* EWALD ELECTROSTATICS */
1171 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1172 ewrt = _mm_mul_pd(r31,ewtabscale);
1173 ewitab = _mm_cvttpd_epi32(ewrt);
1174 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1175 ewitab = _mm_slli_epi32(ewitab,2);
1176 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1177 ewtabD = _mm_setzero_pd();
1178 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1179 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1180 ewtabFn = _mm_setzero_pd();
1181 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1182 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1183 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1184 velec = _mm_mul_pd(qq31,_mm_sub_pd(rinv31,velec));
1185 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1187 /* Update potential sum for this i atom from the interaction with this j atom. */
1188 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1189 velecsum = _mm_add_pd(velecsum,velec);
1193 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1195 /* Calculate temporary vectorial force */
1196 tx = _mm_mul_pd(fscal,dx31);
1197 ty = _mm_mul_pd(fscal,dy31);
1198 tz = _mm_mul_pd(fscal,dz31);
1200 /* Update vectorial force */
1201 fix3 = _mm_add_pd(fix3,tx);
1202 fiy3 = _mm_add_pd(fiy3,ty);
1203 fiz3 = _mm_add_pd(fiz3,tz);
1205 fjx1 = _mm_add_pd(fjx1,tx);
1206 fjy1 = _mm_add_pd(fjy1,ty);
1207 fjz1 = _mm_add_pd(fjz1,tz);
1209 /**************************
1210 * CALCULATE INTERACTIONS *
1211 **************************/
1213 r32 = _mm_mul_pd(rsq32,rinv32);
1215 /* EWALD ELECTROSTATICS */
1217 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1218 ewrt = _mm_mul_pd(r32,ewtabscale);
1219 ewitab = _mm_cvttpd_epi32(ewrt);
1220 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1221 ewitab = _mm_slli_epi32(ewitab,2);
1222 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1223 ewtabD = _mm_setzero_pd();
1224 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1225 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1226 ewtabFn = _mm_setzero_pd();
1227 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1228 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1229 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1230 velec = _mm_mul_pd(qq32,_mm_sub_pd(rinv32,velec));
1231 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1233 /* Update potential sum for this i atom from the interaction with this j atom. */
1234 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1235 velecsum = _mm_add_pd(velecsum,velec);
1239 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1241 /* Calculate temporary vectorial force */
1242 tx = _mm_mul_pd(fscal,dx32);
1243 ty = _mm_mul_pd(fscal,dy32);
1244 tz = _mm_mul_pd(fscal,dz32);
1246 /* Update vectorial force */
1247 fix3 = _mm_add_pd(fix3,tx);
1248 fiy3 = _mm_add_pd(fiy3,ty);
1249 fiz3 = _mm_add_pd(fiz3,tz);
1251 fjx2 = _mm_add_pd(fjx2,tx);
1252 fjy2 = _mm_add_pd(fjy2,ty);
1253 fjz2 = _mm_add_pd(fjz2,tz);
1255 /**************************
1256 * CALCULATE INTERACTIONS *
1257 **************************/
1259 r33 = _mm_mul_pd(rsq33,rinv33);
1261 /* EWALD ELECTROSTATICS */
1263 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1264 ewrt = _mm_mul_pd(r33,ewtabscale);
1265 ewitab = _mm_cvttpd_epi32(ewrt);
1266 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1267 ewitab = _mm_slli_epi32(ewitab,2);
1268 ewtabF = _mm_load_pd( ewtab + gmx_mm_extract_epi32(ewitab,0) );
1269 ewtabD = _mm_setzero_pd();
1270 GMX_MM_TRANSPOSE2_PD(ewtabF,ewtabD);
1271 ewtabV = _mm_load_sd( ewtab + gmx_mm_extract_epi32(ewitab,0) +2);
1272 ewtabFn = _mm_setzero_pd();
1273 GMX_MM_TRANSPOSE2_PD(ewtabV,ewtabFn);
1274 felec = _mm_add_pd(ewtabF,_mm_mul_pd(eweps,ewtabD));
1275 velec = _mm_sub_pd(ewtabV,_mm_mul_pd(_mm_mul_pd(ewtabhalfspace,eweps),_mm_add_pd(ewtabF,felec)));
1276 velec = _mm_mul_pd(qq33,_mm_sub_pd(rinv33,velec));
1277 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1279 /* Update potential sum for this i atom from the interaction with this j atom. */
1280 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
1281 velecsum = _mm_add_pd(velecsum,velec);
1285 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
1287 /* Calculate temporary vectorial force */
1288 tx = _mm_mul_pd(fscal,dx33);
1289 ty = _mm_mul_pd(fscal,dy33);
1290 tz = _mm_mul_pd(fscal,dz33);
1292 /* Update vectorial force */
1293 fix3 = _mm_add_pd(fix3,tx);
1294 fiy3 = _mm_add_pd(fiy3,ty);
1295 fiz3 = _mm_add_pd(fiz3,tz);
1297 fjx3 = _mm_add_pd(fjx3,tx);
1298 fjy3 = _mm_add_pd(fjy3,ty);
1299 fjz3 = _mm_add_pd(fjz3,tz);
1301 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1303 /* Inner loop uses 423 flops */
1306 /* End of innermost loop */
1308 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1309 f+i_coord_offset,fshift+i_shift_offset);
1312 /* Update potential energies */
1313 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
1314 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
1316 /* Increment number of inner iterations */
1317 inneriter += j_index_end - j_index_start;
1319 /* Outer loop uses 26 flops */
1322 /* Increment number of outer iterations */
1325 /* Update outer/inner flops */
1327 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_VF,outeriter*26 + inneriter*423);
1330 * Gromacs nonbonded kernel: nb_kernel_ElecEw_VdwLJEw_GeomW4W4_F_sse2_double
1331 * Electrostatics interaction: Ewald
1332 * VdW interaction: LJEwald
1333 * Geometry: Water4-Water4
1334 * Calculate force/pot: Force
1337 nb_kernel_ElecEw_VdwLJEw_GeomW4W4_F_sse2_double
1338 (t_nblist * gmx_restrict nlist,
1339 rvec * gmx_restrict xx,
1340 rvec * gmx_restrict ff,
1341 t_forcerec * gmx_restrict fr,
1342 t_mdatoms * gmx_restrict mdatoms,
1343 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
1344 t_nrnb * gmx_restrict nrnb)
1346 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
1347 * just 0 for non-waters.
1348 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
1349 * jnr indices corresponding to data put in the four positions in the SIMD register.
1351 int i_shift_offset,i_coord_offset,outeriter,inneriter;
1352 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
1354 int j_coord_offsetA,j_coord_offsetB;
1355 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
1356 real rcutoff_scalar;
1357 real *shiftvec,*fshift,*x,*f;
1358 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
1360 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
1362 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
1364 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
1366 __m128d ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
1367 int vdwjidx0A,vdwjidx0B;
1368 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
1369 int vdwjidx1A,vdwjidx1B;
1370 __m128d jx1,jy1,jz1,fjx1,fjy1,fjz1,jq1,isaj1;
1371 int vdwjidx2A,vdwjidx2B;
1372 __m128d jx2,jy2,jz2,fjx2,fjy2,fjz2,jq2,isaj2;
1373 int vdwjidx3A,vdwjidx3B;
1374 __m128d jx3,jy3,jz3,fjx3,fjy3,fjz3,jq3,isaj3;
1375 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
1376 __m128d dx11,dy11,dz11,rsq11,rinv11,rinvsq11,r11,qq11,c6_11,c12_11;
1377 __m128d dx12,dy12,dz12,rsq12,rinv12,rinvsq12,r12,qq12,c6_12,c12_12;
1378 __m128d dx13,dy13,dz13,rsq13,rinv13,rinvsq13,r13,qq13,c6_13,c12_13;
1379 __m128d dx21,dy21,dz21,rsq21,rinv21,rinvsq21,r21,qq21,c6_21,c12_21;
1380 __m128d dx22,dy22,dz22,rsq22,rinv22,rinvsq22,r22,qq22,c6_22,c12_22;
1381 __m128d dx23,dy23,dz23,rsq23,rinv23,rinvsq23,r23,qq23,c6_23,c12_23;
1382 __m128d dx31,dy31,dz31,rsq31,rinv31,rinvsq31,r31,qq31,c6_31,c12_31;
1383 __m128d dx32,dy32,dz32,rsq32,rinv32,rinvsq32,r32,qq32,c6_32,c12_32;
1384 __m128d dx33,dy33,dz33,rsq33,rinv33,rinvsq33,r33,qq33,c6_33,c12_33;
1385 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
1388 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
1391 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
1392 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
1403 __m128d ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
1405 __m128d one_half = _mm_set1_pd(0.5);
1406 __m128d minus_one = _mm_set1_pd(-1.0);
1408 __m128d ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
1410 __m128d dummy_mask,cutoff_mask;
1411 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
1412 __m128d one = _mm_set1_pd(1.0);
1413 __m128d two = _mm_set1_pd(2.0);
1419 jindex = nlist->jindex;
1421 shiftidx = nlist->shift;
1423 shiftvec = fr->shift_vec[0];
1424 fshift = fr->fshift[0];
1425 facel = _mm_set1_pd(fr->epsfac);
1426 charge = mdatoms->chargeA;
1427 nvdwtype = fr->ntype;
1428 vdwparam = fr->nbfp;
1429 vdwtype = mdatoms->typeA;
1430 vdwgridparam = fr->ljpme_c6grid;
1431 sh_lj_ewald = _mm_set1_pd(fr->ic->sh_lj_ewald);
1432 ewclj = _mm_set1_pd(fr->ewaldcoeff_lj);
1433 ewclj2 = _mm_mul_pd(minus_one,_mm_mul_pd(ewclj,ewclj));
1435 sh_ewald = _mm_set1_pd(fr->ic->sh_ewald);
1436 ewtab = fr->ic->tabq_coul_F;
1437 ewtabscale = _mm_set1_pd(fr->ic->tabq_scale);
1438 ewtabhalfspace = _mm_set1_pd(0.5/fr->ic->tabq_scale);
1440 /* Setup water-specific parameters */
1441 inr = nlist->iinr[0];
1442 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
1443 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
1444 iq3 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+3]));
1445 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
1447 jq1 = _mm_set1_pd(charge[inr+1]);
1448 jq2 = _mm_set1_pd(charge[inr+2]);
1449 jq3 = _mm_set1_pd(charge[inr+3]);
1450 vdwjidx0A = 2*vdwtype[inr+0];
1451 c6_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A]);
1452 c12_00 = _mm_set1_pd(vdwparam[vdwioffset0+vdwjidx0A+1]);
1453 c6grid_00 = _mm_set1_pd(vdwgridparam[vdwioffset0+vdwjidx0A]);
1454 qq11 = _mm_mul_pd(iq1,jq1);
1455 qq12 = _mm_mul_pd(iq1,jq2);
1456 qq13 = _mm_mul_pd(iq1,jq3);
1457 qq21 = _mm_mul_pd(iq2,jq1);
1458 qq22 = _mm_mul_pd(iq2,jq2);
1459 qq23 = _mm_mul_pd(iq2,jq3);
1460 qq31 = _mm_mul_pd(iq3,jq1);
1461 qq32 = _mm_mul_pd(iq3,jq2);
1462 qq33 = _mm_mul_pd(iq3,jq3);
1464 /* Avoid stupid compiler warnings */
1466 j_coord_offsetA = 0;
1467 j_coord_offsetB = 0;
1472 /* Start outer loop over neighborlists */
1473 for(iidx=0; iidx<nri; iidx++)
1475 /* Load shift vector for this list */
1476 i_shift_offset = DIM*shiftidx[iidx];
1478 /* Load limits for loop over neighbors */
1479 j_index_start = jindex[iidx];
1480 j_index_end = jindex[iidx+1];
1482 /* Get outer coordinate index */
1484 i_coord_offset = DIM*inr;
1486 /* Load i particle coords and add shift vector */
1487 gmx_mm_load_shift_and_4rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
1488 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
1490 fix0 = _mm_setzero_pd();
1491 fiy0 = _mm_setzero_pd();
1492 fiz0 = _mm_setzero_pd();
1493 fix1 = _mm_setzero_pd();
1494 fiy1 = _mm_setzero_pd();
1495 fiz1 = _mm_setzero_pd();
1496 fix2 = _mm_setzero_pd();
1497 fiy2 = _mm_setzero_pd();
1498 fiz2 = _mm_setzero_pd();
1499 fix3 = _mm_setzero_pd();
1500 fiy3 = _mm_setzero_pd();
1501 fiz3 = _mm_setzero_pd();
1503 /* Start inner kernel loop */
1504 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
1507 /* Get j neighbor index, and coordinate index */
1509 jnrB = jjnr[jidx+1];
1510 j_coord_offsetA = DIM*jnrA;
1511 j_coord_offsetB = DIM*jnrB;
1513 /* load j atom coordinates */
1514 gmx_mm_load_4rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
1515 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1516 &jy2,&jz2,&jx3,&jy3,&jz3);
1518 /* Calculate displacement vector */
1519 dx00 = _mm_sub_pd(ix0,jx0);
1520 dy00 = _mm_sub_pd(iy0,jy0);
1521 dz00 = _mm_sub_pd(iz0,jz0);
1522 dx11 = _mm_sub_pd(ix1,jx1);
1523 dy11 = _mm_sub_pd(iy1,jy1);
1524 dz11 = _mm_sub_pd(iz1,jz1);
1525 dx12 = _mm_sub_pd(ix1,jx2);
1526 dy12 = _mm_sub_pd(iy1,jy2);
1527 dz12 = _mm_sub_pd(iz1,jz2);
1528 dx13 = _mm_sub_pd(ix1,jx3);
1529 dy13 = _mm_sub_pd(iy1,jy3);
1530 dz13 = _mm_sub_pd(iz1,jz3);
1531 dx21 = _mm_sub_pd(ix2,jx1);
1532 dy21 = _mm_sub_pd(iy2,jy1);
1533 dz21 = _mm_sub_pd(iz2,jz1);
1534 dx22 = _mm_sub_pd(ix2,jx2);
1535 dy22 = _mm_sub_pd(iy2,jy2);
1536 dz22 = _mm_sub_pd(iz2,jz2);
1537 dx23 = _mm_sub_pd(ix2,jx3);
1538 dy23 = _mm_sub_pd(iy2,jy3);
1539 dz23 = _mm_sub_pd(iz2,jz3);
1540 dx31 = _mm_sub_pd(ix3,jx1);
1541 dy31 = _mm_sub_pd(iy3,jy1);
1542 dz31 = _mm_sub_pd(iz3,jz1);
1543 dx32 = _mm_sub_pd(ix3,jx2);
1544 dy32 = _mm_sub_pd(iy3,jy2);
1545 dz32 = _mm_sub_pd(iz3,jz2);
1546 dx33 = _mm_sub_pd(ix3,jx3);
1547 dy33 = _mm_sub_pd(iy3,jy3);
1548 dz33 = _mm_sub_pd(iz3,jz3);
1550 /* Calculate squared distance and things based on it */
1551 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1552 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1553 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1554 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1555 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1556 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1557 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1558 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1559 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1560 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1562 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1563 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1564 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1565 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1566 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1567 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1568 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1569 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1570 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1571 rinv33 = gmx_mm_invsqrt_pd(rsq33);
1573 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
1574 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
1575 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
1576 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
1577 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
1578 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
1579 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
1580 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
1581 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
1582 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
1584 fjx0 = _mm_setzero_pd();
1585 fjy0 = _mm_setzero_pd();
1586 fjz0 = _mm_setzero_pd();
1587 fjx1 = _mm_setzero_pd();
1588 fjy1 = _mm_setzero_pd();
1589 fjz1 = _mm_setzero_pd();
1590 fjx2 = _mm_setzero_pd();
1591 fjy2 = _mm_setzero_pd();
1592 fjz2 = _mm_setzero_pd();
1593 fjx3 = _mm_setzero_pd();
1594 fjy3 = _mm_setzero_pd();
1595 fjz3 = _mm_setzero_pd();
1597 /**************************
1598 * CALCULATE INTERACTIONS *
1599 **************************/
1601 r00 = _mm_mul_pd(rsq00,rinv00);
1603 /* Analytical LJ-PME */
1604 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
1605 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
1606 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
1607 exponent = gmx_simd_exp_d(ewcljrsq);
1608 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1609 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
1610 /* f6A = 6 * C6grid * (1 - poly) */
1611 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
1612 /* f6B = C6grid * exponent * beta^6 */
1613 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
1614 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1615 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
1619 /* Calculate temporary vectorial force */
1620 tx = _mm_mul_pd(fscal,dx00);
1621 ty = _mm_mul_pd(fscal,dy00);
1622 tz = _mm_mul_pd(fscal,dz00);
1624 /* Update vectorial force */
1625 fix0 = _mm_add_pd(fix0,tx);
1626 fiy0 = _mm_add_pd(fiy0,ty);
1627 fiz0 = _mm_add_pd(fiz0,tz);
1629 fjx0 = _mm_add_pd(fjx0,tx);
1630 fjy0 = _mm_add_pd(fjy0,ty);
1631 fjz0 = _mm_add_pd(fjz0,tz);
1633 /**************************
1634 * CALCULATE INTERACTIONS *
1635 **************************/
1637 r11 = _mm_mul_pd(rsq11,rinv11);
1639 /* EWALD ELECTROSTATICS */
1641 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1642 ewrt = _mm_mul_pd(r11,ewtabscale);
1643 ewitab = _mm_cvttpd_epi32(ewrt);
1644 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1645 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1647 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1648 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
1652 /* Calculate temporary vectorial force */
1653 tx = _mm_mul_pd(fscal,dx11);
1654 ty = _mm_mul_pd(fscal,dy11);
1655 tz = _mm_mul_pd(fscal,dz11);
1657 /* Update vectorial force */
1658 fix1 = _mm_add_pd(fix1,tx);
1659 fiy1 = _mm_add_pd(fiy1,ty);
1660 fiz1 = _mm_add_pd(fiz1,tz);
1662 fjx1 = _mm_add_pd(fjx1,tx);
1663 fjy1 = _mm_add_pd(fjy1,ty);
1664 fjz1 = _mm_add_pd(fjz1,tz);
1666 /**************************
1667 * CALCULATE INTERACTIONS *
1668 **************************/
1670 r12 = _mm_mul_pd(rsq12,rinv12);
1672 /* EWALD ELECTROSTATICS */
1674 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1675 ewrt = _mm_mul_pd(r12,ewtabscale);
1676 ewitab = _mm_cvttpd_epi32(ewrt);
1677 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1678 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1680 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1681 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
1685 /* Calculate temporary vectorial force */
1686 tx = _mm_mul_pd(fscal,dx12);
1687 ty = _mm_mul_pd(fscal,dy12);
1688 tz = _mm_mul_pd(fscal,dz12);
1690 /* Update vectorial force */
1691 fix1 = _mm_add_pd(fix1,tx);
1692 fiy1 = _mm_add_pd(fiy1,ty);
1693 fiz1 = _mm_add_pd(fiz1,tz);
1695 fjx2 = _mm_add_pd(fjx2,tx);
1696 fjy2 = _mm_add_pd(fjy2,ty);
1697 fjz2 = _mm_add_pd(fjz2,tz);
1699 /**************************
1700 * CALCULATE INTERACTIONS *
1701 **************************/
1703 r13 = _mm_mul_pd(rsq13,rinv13);
1705 /* EWALD ELECTROSTATICS */
1707 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1708 ewrt = _mm_mul_pd(r13,ewtabscale);
1709 ewitab = _mm_cvttpd_epi32(ewrt);
1710 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1711 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1713 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1714 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
1718 /* Calculate temporary vectorial force */
1719 tx = _mm_mul_pd(fscal,dx13);
1720 ty = _mm_mul_pd(fscal,dy13);
1721 tz = _mm_mul_pd(fscal,dz13);
1723 /* Update vectorial force */
1724 fix1 = _mm_add_pd(fix1,tx);
1725 fiy1 = _mm_add_pd(fiy1,ty);
1726 fiz1 = _mm_add_pd(fiz1,tz);
1728 fjx3 = _mm_add_pd(fjx3,tx);
1729 fjy3 = _mm_add_pd(fjy3,ty);
1730 fjz3 = _mm_add_pd(fjz3,tz);
1732 /**************************
1733 * CALCULATE INTERACTIONS *
1734 **************************/
1736 r21 = _mm_mul_pd(rsq21,rinv21);
1738 /* EWALD ELECTROSTATICS */
1740 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1741 ewrt = _mm_mul_pd(r21,ewtabscale);
1742 ewitab = _mm_cvttpd_epi32(ewrt);
1743 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1744 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1746 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1747 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
1751 /* Calculate temporary vectorial force */
1752 tx = _mm_mul_pd(fscal,dx21);
1753 ty = _mm_mul_pd(fscal,dy21);
1754 tz = _mm_mul_pd(fscal,dz21);
1756 /* Update vectorial force */
1757 fix2 = _mm_add_pd(fix2,tx);
1758 fiy2 = _mm_add_pd(fiy2,ty);
1759 fiz2 = _mm_add_pd(fiz2,tz);
1761 fjx1 = _mm_add_pd(fjx1,tx);
1762 fjy1 = _mm_add_pd(fjy1,ty);
1763 fjz1 = _mm_add_pd(fjz1,tz);
1765 /**************************
1766 * CALCULATE INTERACTIONS *
1767 **************************/
1769 r22 = _mm_mul_pd(rsq22,rinv22);
1771 /* EWALD ELECTROSTATICS */
1773 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1774 ewrt = _mm_mul_pd(r22,ewtabscale);
1775 ewitab = _mm_cvttpd_epi32(ewrt);
1776 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1777 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1779 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1780 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
1784 /* Calculate temporary vectorial force */
1785 tx = _mm_mul_pd(fscal,dx22);
1786 ty = _mm_mul_pd(fscal,dy22);
1787 tz = _mm_mul_pd(fscal,dz22);
1789 /* Update vectorial force */
1790 fix2 = _mm_add_pd(fix2,tx);
1791 fiy2 = _mm_add_pd(fiy2,ty);
1792 fiz2 = _mm_add_pd(fiz2,tz);
1794 fjx2 = _mm_add_pd(fjx2,tx);
1795 fjy2 = _mm_add_pd(fjy2,ty);
1796 fjz2 = _mm_add_pd(fjz2,tz);
1798 /**************************
1799 * CALCULATE INTERACTIONS *
1800 **************************/
1802 r23 = _mm_mul_pd(rsq23,rinv23);
1804 /* EWALD ELECTROSTATICS */
1806 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1807 ewrt = _mm_mul_pd(r23,ewtabscale);
1808 ewitab = _mm_cvttpd_epi32(ewrt);
1809 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1810 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1812 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1813 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
1817 /* Calculate temporary vectorial force */
1818 tx = _mm_mul_pd(fscal,dx23);
1819 ty = _mm_mul_pd(fscal,dy23);
1820 tz = _mm_mul_pd(fscal,dz23);
1822 /* Update vectorial force */
1823 fix2 = _mm_add_pd(fix2,tx);
1824 fiy2 = _mm_add_pd(fiy2,ty);
1825 fiz2 = _mm_add_pd(fiz2,tz);
1827 fjx3 = _mm_add_pd(fjx3,tx);
1828 fjy3 = _mm_add_pd(fjy3,ty);
1829 fjz3 = _mm_add_pd(fjz3,tz);
1831 /**************************
1832 * CALCULATE INTERACTIONS *
1833 **************************/
1835 r31 = _mm_mul_pd(rsq31,rinv31);
1837 /* EWALD ELECTROSTATICS */
1839 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1840 ewrt = _mm_mul_pd(r31,ewtabscale);
1841 ewitab = _mm_cvttpd_epi32(ewrt);
1842 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1843 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1845 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1846 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
1850 /* Calculate temporary vectorial force */
1851 tx = _mm_mul_pd(fscal,dx31);
1852 ty = _mm_mul_pd(fscal,dy31);
1853 tz = _mm_mul_pd(fscal,dz31);
1855 /* Update vectorial force */
1856 fix3 = _mm_add_pd(fix3,tx);
1857 fiy3 = _mm_add_pd(fiy3,ty);
1858 fiz3 = _mm_add_pd(fiz3,tz);
1860 fjx1 = _mm_add_pd(fjx1,tx);
1861 fjy1 = _mm_add_pd(fjy1,ty);
1862 fjz1 = _mm_add_pd(fjz1,tz);
1864 /**************************
1865 * CALCULATE INTERACTIONS *
1866 **************************/
1868 r32 = _mm_mul_pd(rsq32,rinv32);
1870 /* EWALD ELECTROSTATICS */
1872 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1873 ewrt = _mm_mul_pd(r32,ewtabscale);
1874 ewitab = _mm_cvttpd_epi32(ewrt);
1875 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1876 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1878 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1879 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
1883 /* Calculate temporary vectorial force */
1884 tx = _mm_mul_pd(fscal,dx32);
1885 ty = _mm_mul_pd(fscal,dy32);
1886 tz = _mm_mul_pd(fscal,dz32);
1888 /* Update vectorial force */
1889 fix3 = _mm_add_pd(fix3,tx);
1890 fiy3 = _mm_add_pd(fiy3,ty);
1891 fiz3 = _mm_add_pd(fiz3,tz);
1893 fjx2 = _mm_add_pd(fjx2,tx);
1894 fjy2 = _mm_add_pd(fjy2,ty);
1895 fjz2 = _mm_add_pd(fjz2,tz);
1897 /**************************
1898 * CALCULATE INTERACTIONS *
1899 **************************/
1901 r33 = _mm_mul_pd(rsq33,rinv33);
1903 /* EWALD ELECTROSTATICS */
1905 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1906 ewrt = _mm_mul_pd(r33,ewtabscale);
1907 ewitab = _mm_cvttpd_epi32(ewrt);
1908 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
1909 gmx_mm_load_2pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),ewtab+gmx_mm_extract_epi32(ewitab,1),
1911 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
1912 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
1916 /* Calculate temporary vectorial force */
1917 tx = _mm_mul_pd(fscal,dx33);
1918 ty = _mm_mul_pd(fscal,dy33);
1919 tz = _mm_mul_pd(fscal,dz33);
1921 /* Update vectorial force */
1922 fix3 = _mm_add_pd(fix3,tx);
1923 fiy3 = _mm_add_pd(fiy3,ty);
1924 fiz3 = _mm_add_pd(fiz3,tz);
1926 fjx3 = _mm_add_pd(fjx3,tx);
1927 fjy3 = _mm_add_pd(fjy3,ty);
1928 fjz3 = _mm_add_pd(fjz3,tz);
1930 gmx_mm_decrement_4rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
1932 /* Inner loop uses 373 flops */
1935 if(jidx<j_index_end)
1939 j_coord_offsetA = DIM*jnrA;
1941 /* load j atom coordinates */
1942 gmx_mm_load_4rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
1943 &jx0,&jy0,&jz0,&jx1,&jy1,&jz1,&jx2,
1944 &jy2,&jz2,&jx3,&jy3,&jz3);
1946 /* Calculate displacement vector */
1947 dx00 = _mm_sub_pd(ix0,jx0);
1948 dy00 = _mm_sub_pd(iy0,jy0);
1949 dz00 = _mm_sub_pd(iz0,jz0);
1950 dx11 = _mm_sub_pd(ix1,jx1);
1951 dy11 = _mm_sub_pd(iy1,jy1);
1952 dz11 = _mm_sub_pd(iz1,jz1);
1953 dx12 = _mm_sub_pd(ix1,jx2);
1954 dy12 = _mm_sub_pd(iy1,jy2);
1955 dz12 = _mm_sub_pd(iz1,jz2);
1956 dx13 = _mm_sub_pd(ix1,jx3);
1957 dy13 = _mm_sub_pd(iy1,jy3);
1958 dz13 = _mm_sub_pd(iz1,jz3);
1959 dx21 = _mm_sub_pd(ix2,jx1);
1960 dy21 = _mm_sub_pd(iy2,jy1);
1961 dz21 = _mm_sub_pd(iz2,jz1);
1962 dx22 = _mm_sub_pd(ix2,jx2);
1963 dy22 = _mm_sub_pd(iy2,jy2);
1964 dz22 = _mm_sub_pd(iz2,jz2);
1965 dx23 = _mm_sub_pd(ix2,jx3);
1966 dy23 = _mm_sub_pd(iy2,jy3);
1967 dz23 = _mm_sub_pd(iz2,jz3);
1968 dx31 = _mm_sub_pd(ix3,jx1);
1969 dy31 = _mm_sub_pd(iy3,jy1);
1970 dz31 = _mm_sub_pd(iz3,jz1);
1971 dx32 = _mm_sub_pd(ix3,jx2);
1972 dy32 = _mm_sub_pd(iy3,jy2);
1973 dz32 = _mm_sub_pd(iz3,jz2);
1974 dx33 = _mm_sub_pd(ix3,jx3);
1975 dy33 = _mm_sub_pd(iy3,jy3);
1976 dz33 = _mm_sub_pd(iz3,jz3);
1978 /* Calculate squared distance and things based on it */
1979 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
1980 rsq11 = gmx_mm_calc_rsq_pd(dx11,dy11,dz11);
1981 rsq12 = gmx_mm_calc_rsq_pd(dx12,dy12,dz12);
1982 rsq13 = gmx_mm_calc_rsq_pd(dx13,dy13,dz13);
1983 rsq21 = gmx_mm_calc_rsq_pd(dx21,dy21,dz21);
1984 rsq22 = gmx_mm_calc_rsq_pd(dx22,dy22,dz22);
1985 rsq23 = gmx_mm_calc_rsq_pd(dx23,dy23,dz23);
1986 rsq31 = gmx_mm_calc_rsq_pd(dx31,dy31,dz31);
1987 rsq32 = gmx_mm_calc_rsq_pd(dx32,dy32,dz32);
1988 rsq33 = gmx_mm_calc_rsq_pd(dx33,dy33,dz33);
1990 rinv00 = gmx_mm_invsqrt_pd(rsq00);
1991 rinv11 = gmx_mm_invsqrt_pd(rsq11);
1992 rinv12 = gmx_mm_invsqrt_pd(rsq12);
1993 rinv13 = gmx_mm_invsqrt_pd(rsq13);
1994 rinv21 = gmx_mm_invsqrt_pd(rsq21);
1995 rinv22 = gmx_mm_invsqrt_pd(rsq22);
1996 rinv23 = gmx_mm_invsqrt_pd(rsq23);
1997 rinv31 = gmx_mm_invsqrt_pd(rsq31);
1998 rinv32 = gmx_mm_invsqrt_pd(rsq32);
1999 rinv33 = gmx_mm_invsqrt_pd(rsq33);
2001 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
2002 rinvsq11 = _mm_mul_pd(rinv11,rinv11);
2003 rinvsq12 = _mm_mul_pd(rinv12,rinv12);
2004 rinvsq13 = _mm_mul_pd(rinv13,rinv13);
2005 rinvsq21 = _mm_mul_pd(rinv21,rinv21);
2006 rinvsq22 = _mm_mul_pd(rinv22,rinv22);
2007 rinvsq23 = _mm_mul_pd(rinv23,rinv23);
2008 rinvsq31 = _mm_mul_pd(rinv31,rinv31);
2009 rinvsq32 = _mm_mul_pd(rinv32,rinv32);
2010 rinvsq33 = _mm_mul_pd(rinv33,rinv33);
2012 fjx0 = _mm_setzero_pd();
2013 fjy0 = _mm_setzero_pd();
2014 fjz0 = _mm_setzero_pd();
2015 fjx1 = _mm_setzero_pd();
2016 fjy1 = _mm_setzero_pd();
2017 fjz1 = _mm_setzero_pd();
2018 fjx2 = _mm_setzero_pd();
2019 fjy2 = _mm_setzero_pd();
2020 fjz2 = _mm_setzero_pd();
2021 fjx3 = _mm_setzero_pd();
2022 fjy3 = _mm_setzero_pd();
2023 fjz3 = _mm_setzero_pd();
2025 /**************************
2026 * CALCULATE INTERACTIONS *
2027 **************************/
2029 r00 = _mm_mul_pd(rsq00,rinv00);
2031 /* Analytical LJ-PME */
2032 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
2033 ewcljrsq = _mm_mul_pd(ewclj2,rsq00);
2034 ewclj6 = _mm_mul_pd(ewclj2,_mm_mul_pd(ewclj2,ewclj2));
2035 exponent = gmx_simd_exp_d(ewcljrsq);
2036 /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
2037 poly = _mm_mul_pd(exponent,_mm_add_pd(_mm_sub_pd(one,ewcljrsq),_mm_mul_pd(_mm_mul_pd(ewcljrsq,ewcljrsq),one_half)));
2038 /* f6A = 6 * C6grid * (1 - poly) */
2039 f6A = _mm_mul_pd(c6grid_00,_mm_sub_pd(one,poly));
2040 /* f6B = C6grid * exponent * beta^6 */
2041 f6B = _mm_mul_pd(_mm_mul_pd(c6grid_00,one_sixth),_mm_mul_pd(exponent,ewclj6));
2042 /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
2043 fvdw = _mm_mul_pd(_mm_add_pd(_mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),_mm_sub_pd(c6_00,f6A)),rinvsix),f6B),rinvsq00);
2047 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2049 /* Calculate temporary vectorial force */
2050 tx = _mm_mul_pd(fscal,dx00);
2051 ty = _mm_mul_pd(fscal,dy00);
2052 tz = _mm_mul_pd(fscal,dz00);
2054 /* Update vectorial force */
2055 fix0 = _mm_add_pd(fix0,tx);
2056 fiy0 = _mm_add_pd(fiy0,ty);
2057 fiz0 = _mm_add_pd(fiz0,tz);
2059 fjx0 = _mm_add_pd(fjx0,tx);
2060 fjy0 = _mm_add_pd(fjy0,ty);
2061 fjz0 = _mm_add_pd(fjz0,tz);
2063 /**************************
2064 * CALCULATE INTERACTIONS *
2065 **************************/
2067 r11 = _mm_mul_pd(rsq11,rinv11);
2069 /* EWALD ELECTROSTATICS */
2071 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2072 ewrt = _mm_mul_pd(r11,ewtabscale);
2073 ewitab = _mm_cvttpd_epi32(ewrt);
2074 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2075 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2076 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2077 felec = _mm_mul_pd(_mm_mul_pd(qq11,rinv11),_mm_sub_pd(rinvsq11,felec));
2081 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2083 /* Calculate temporary vectorial force */
2084 tx = _mm_mul_pd(fscal,dx11);
2085 ty = _mm_mul_pd(fscal,dy11);
2086 tz = _mm_mul_pd(fscal,dz11);
2088 /* Update vectorial force */
2089 fix1 = _mm_add_pd(fix1,tx);
2090 fiy1 = _mm_add_pd(fiy1,ty);
2091 fiz1 = _mm_add_pd(fiz1,tz);
2093 fjx1 = _mm_add_pd(fjx1,tx);
2094 fjy1 = _mm_add_pd(fjy1,ty);
2095 fjz1 = _mm_add_pd(fjz1,tz);
2097 /**************************
2098 * CALCULATE INTERACTIONS *
2099 **************************/
2101 r12 = _mm_mul_pd(rsq12,rinv12);
2103 /* EWALD ELECTROSTATICS */
2105 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2106 ewrt = _mm_mul_pd(r12,ewtabscale);
2107 ewitab = _mm_cvttpd_epi32(ewrt);
2108 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2109 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2110 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2111 felec = _mm_mul_pd(_mm_mul_pd(qq12,rinv12),_mm_sub_pd(rinvsq12,felec));
2115 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2117 /* Calculate temporary vectorial force */
2118 tx = _mm_mul_pd(fscal,dx12);
2119 ty = _mm_mul_pd(fscal,dy12);
2120 tz = _mm_mul_pd(fscal,dz12);
2122 /* Update vectorial force */
2123 fix1 = _mm_add_pd(fix1,tx);
2124 fiy1 = _mm_add_pd(fiy1,ty);
2125 fiz1 = _mm_add_pd(fiz1,tz);
2127 fjx2 = _mm_add_pd(fjx2,tx);
2128 fjy2 = _mm_add_pd(fjy2,ty);
2129 fjz2 = _mm_add_pd(fjz2,tz);
2131 /**************************
2132 * CALCULATE INTERACTIONS *
2133 **************************/
2135 r13 = _mm_mul_pd(rsq13,rinv13);
2137 /* EWALD ELECTROSTATICS */
2139 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2140 ewrt = _mm_mul_pd(r13,ewtabscale);
2141 ewitab = _mm_cvttpd_epi32(ewrt);
2142 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2143 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2144 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2145 felec = _mm_mul_pd(_mm_mul_pd(qq13,rinv13),_mm_sub_pd(rinvsq13,felec));
2149 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2151 /* Calculate temporary vectorial force */
2152 tx = _mm_mul_pd(fscal,dx13);
2153 ty = _mm_mul_pd(fscal,dy13);
2154 tz = _mm_mul_pd(fscal,dz13);
2156 /* Update vectorial force */
2157 fix1 = _mm_add_pd(fix1,tx);
2158 fiy1 = _mm_add_pd(fiy1,ty);
2159 fiz1 = _mm_add_pd(fiz1,tz);
2161 fjx3 = _mm_add_pd(fjx3,tx);
2162 fjy3 = _mm_add_pd(fjy3,ty);
2163 fjz3 = _mm_add_pd(fjz3,tz);
2165 /**************************
2166 * CALCULATE INTERACTIONS *
2167 **************************/
2169 r21 = _mm_mul_pd(rsq21,rinv21);
2171 /* EWALD ELECTROSTATICS */
2173 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2174 ewrt = _mm_mul_pd(r21,ewtabscale);
2175 ewitab = _mm_cvttpd_epi32(ewrt);
2176 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2177 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2178 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2179 felec = _mm_mul_pd(_mm_mul_pd(qq21,rinv21),_mm_sub_pd(rinvsq21,felec));
2183 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2185 /* Calculate temporary vectorial force */
2186 tx = _mm_mul_pd(fscal,dx21);
2187 ty = _mm_mul_pd(fscal,dy21);
2188 tz = _mm_mul_pd(fscal,dz21);
2190 /* Update vectorial force */
2191 fix2 = _mm_add_pd(fix2,tx);
2192 fiy2 = _mm_add_pd(fiy2,ty);
2193 fiz2 = _mm_add_pd(fiz2,tz);
2195 fjx1 = _mm_add_pd(fjx1,tx);
2196 fjy1 = _mm_add_pd(fjy1,ty);
2197 fjz1 = _mm_add_pd(fjz1,tz);
2199 /**************************
2200 * CALCULATE INTERACTIONS *
2201 **************************/
2203 r22 = _mm_mul_pd(rsq22,rinv22);
2205 /* EWALD ELECTROSTATICS */
2207 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2208 ewrt = _mm_mul_pd(r22,ewtabscale);
2209 ewitab = _mm_cvttpd_epi32(ewrt);
2210 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2211 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2212 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2213 felec = _mm_mul_pd(_mm_mul_pd(qq22,rinv22),_mm_sub_pd(rinvsq22,felec));
2217 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2219 /* Calculate temporary vectorial force */
2220 tx = _mm_mul_pd(fscal,dx22);
2221 ty = _mm_mul_pd(fscal,dy22);
2222 tz = _mm_mul_pd(fscal,dz22);
2224 /* Update vectorial force */
2225 fix2 = _mm_add_pd(fix2,tx);
2226 fiy2 = _mm_add_pd(fiy2,ty);
2227 fiz2 = _mm_add_pd(fiz2,tz);
2229 fjx2 = _mm_add_pd(fjx2,tx);
2230 fjy2 = _mm_add_pd(fjy2,ty);
2231 fjz2 = _mm_add_pd(fjz2,tz);
2233 /**************************
2234 * CALCULATE INTERACTIONS *
2235 **************************/
2237 r23 = _mm_mul_pd(rsq23,rinv23);
2239 /* EWALD ELECTROSTATICS */
2241 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2242 ewrt = _mm_mul_pd(r23,ewtabscale);
2243 ewitab = _mm_cvttpd_epi32(ewrt);
2244 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2245 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2246 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2247 felec = _mm_mul_pd(_mm_mul_pd(qq23,rinv23),_mm_sub_pd(rinvsq23,felec));
2251 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2253 /* Calculate temporary vectorial force */
2254 tx = _mm_mul_pd(fscal,dx23);
2255 ty = _mm_mul_pd(fscal,dy23);
2256 tz = _mm_mul_pd(fscal,dz23);
2258 /* Update vectorial force */
2259 fix2 = _mm_add_pd(fix2,tx);
2260 fiy2 = _mm_add_pd(fiy2,ty);
2261 fiz2 = _mm_add_pd(fiz2,tz);
2263 fjx3 = _mm_add_pd(fjx3,tx);
2264 fjy3 = _mm_add_pd(fjy3,ty);
2265 fjz3 = _mm_add_pd(fjz3,tz);
2267 /**************************
2268 * CALCULATE INTERACTIONS *
2269 **************************/
2271 r31 = _mm_mul_pd(rsq31,rinv31);
2273 /* EWALD ELECTROSTATICS */
2275 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2276 ewrt = _mm_mul_pd(r31,ewtabscale);
2277 ewitab = _mm_cvttpd_epi32(ewrt);
2278 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2279 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2280 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2281 felec = _mm_mul_pd(_mm_mul_pd(qq31,rinv31),_mm_sub_pd(rinvsq31,felec));
2285 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2287 /* Calculate temporary vectorial force */
2288 tx = _mm_mul_pd(fscal,dx31);
2289 ty = _mm_mul_pd(fscal,dy31);
2290 tz = _mm_mul_pd(fscal,dz31);
2292 /* Update vectorial force */
2293 fix3 = _mm_add_pd(fix3,tx);
2294 fiy3 = _mm_add_pd(fiy3,ty);
2295 fiz3 = _mm_add_pd(fiz3,tz);
2297 fjx1 = _mm_add_pd(fjx1,tx);
2298 fjy1 = _mm_add_pd(fjy1,ty);
2299 fjz1 = _mm_add_pd(fjz1,tz);
2301 /**************************
2302 * CALCULATE INTERACTIONS *
2303 **************************/
2305 r32 = _mm_mul_pd(rsq32,rinv32);
2307 /* EWALD ELECTROSTATICS */
2309 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2310 ewrt = _mm_mul_pd(r32,ewtabscale);
2311 ewitab = _mm_cvttpd_epi32(ewrt);
2312 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2313 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2314 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2315 felec = _mm_mul_pd(_mm_mul_pd(qq32,rinv32),_mm_sub_pd(rinvsq32,felec));
2319 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2321 /* Calculate temporary vectorial force */
2322 tx = _mm_mul_pd(fscal,dx32);
2323 ty = _mm_mul_pd(fscal,dy32);
2324 tz = _mm_mul_pd(fscal,dz32);
2326 /* Update vectorial force */
2327 fix3 = _mm_add_pd(fix3,tx);
2328 fiy3 = _mm_add_pd(fiy3,ty);
2329 fiz3 = _mm_add_pd(fiz3,tz);
2331 fjx2 = _mm_add_pd(fjx2,tx);
2332 fjy2 = _mm_add_pd(fjy2,ty);
2333 fjz2 = _mm_add_pd(fjz2,tz);
2335 /**************************
2336 * CALCULATE INTERACTIONS *
2337 **************************/
2339 r33 = _mm_mul_pd(rsq33,rinv33);
2341 /* EWALD ELECTROSTATICS */
2343 /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
2344 ewrt = _mm_mul_pd(r33,ewtabscale);
2345 ewitab = _mm_cvttpd_epi32(ewrt);
2346 eweps = _mm_sub_pd(ewrt,_mm_cvtepi32_pd(ewitab));
2347 gmx_mm_load_1pair_swizzle_pd(ewtab+gmx_mm_extract_epi32(ewitab,0),&ewtabF,&ewtabFn);
2348 felec = _mm_add_pd(_mm_mul_pd( _mm_sub_pd(one,eweps),ewtabF),_mm_mul_pd(eweps,ewtabFn));
2349 felec = _mm_mul_pd(_mm_mul_pd(qq33,rinv33),_mm_sub_pd(rinvsq33,felec));
2353 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
2355 /* Calculate temporary vectorial force */
2356 tx = _mm_mul_pd(fscal,dx33);
2357 ty = _mm_mul_pd(fscal,dy33);
2358 tz = _mm_mul_pd(fscal,dz33);
2360 /* Update vectorial force */
2361 fix3 = _mm_add_pd(fix3,tx);
2362 fiy3 = _mm_add_pd(fiy3,ty);
2363 fiz3 = _mm_add_pd(fiz3,tz);
2365 fjx3 = _mm_add_pd(fjx3,tx);
2366 fjy3 = _mm_add_pd(fjy3,ty);
2367 fjz3 = _mm_add_pd(fjz3,tz);
2369 gmx_mm_decrement_4rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0,fjx1,fjy1,fjz1,fjx2,fjy2,fjz2,fjx3,fjy3,fjz3);
2371 /* Inner loop uses 373 flops */
2374 /* End of innermost loop */
2376 gmx_mm_update_iforce_4atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
2377 f+i_coord_offset,fshift+i_shift_offset);
2379 /* Increment number of inner iterations */
2380 inneriter += j_index_end - j_index_start;
2382 /* Outer loop uses 24 flops */
2385 /* Increment number of outer iterations */
2388 /* Update outer/inner flops */
2390 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4W4_F,outeriter*24 + inneriter*373);